HAL Id: hal-01251948
https://hal-mines-paristech.archives-ouvertes.fr/hal-01251948
Submitted on 12 Jan 2016
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Mie Equation of State
Alfonso Gonzalez, Luis Pereira, Patrice Paricaud, Christophe Coquelet, Antonin Chapoy
To cite this version:
Alfonso Gonzalez, Luis Pereira, Patrice Paricaud, Christophe Coquelet, Antonin Chapoy. Modeling of
Transport Properties Using the SAFT-VR Mie Equation of State. SPE Annual Technical Conférence
and Exhibition, Sep 2015, HOUSTON, United States. �10.2118/175051-MS�. �hal-01251948�
Modeling of Transport Properties Using the SAFT-VR Mie Equation of State
Alfonso Gonzalez, Heriot-Watt University/Mines-ParisTech PSL CTP, Luis Pereira, Heriot-Watt University, Patrice Paricaud, ENSTA-ParisTech UCP, Christophe Coquelet, Mines-ParisTech PSL CTP, and Antonin Chapoy, Heriot-Watt University
Abstract
Carbon capture and storage (CCS) has been presented as one of the most promising methods to counterbalance the CO
2emissions from the combustion of fossil fuels. Density, viscosity and interfacial tension (IFT) are, among others properties, crucial for the safe and optimum transport and storage of CO
2-rich steams and they play a key role in enhanced oil recovery (EOR) operations. Therefore, in the present work the capability of a new molecular based equation of state (EoS) to describe these properties was evaluated by comparing the model predictions against literature experimental data.
The EoS considered herein is based on an accurate statistical associating fluid theory with variable range interaction through Mie potentials (SAFT-VR Mie EoS). The EoS was used to describe the vapor-liquid equilibria (VLE) and the densities of selected mixtures. Phase equilibrium calculations are then used to estimate viscosity and interfacial tension values. The viscosity model considered is the TRAPP method using the single phase densities, calculated from the EoS. The IFT was evaluated by coupling this EoS with the density gradient theory of fluids interfaces (DGT). The DGT uses bulk phase properties from the mixture to readily estimate the density distribution of each component across the interface and predict interfacial tension values.
To assess the adequacy of the selected models, the modeling results were compared against experimental data of several CO
2-rich systems in a wide range of conditions from the literature. The evaluated systems include five binaries (CO
2/O
2, CO
2/N
2, CO
2/Ar, CO
2/n-C
4and CO
2/n-C
10) and two multicomponent mixtures (90%CO
2/ 5%O
2/ 2%Ar / 3%N
2and 90%CO
2/ 6%n-C
4/ 4%n-C
10).
The modeling results showed low absolute average deviations to the experimental viscosity and IFT data from the literature, supporting the capabilities of the adopted models for describing thermophysical properties of CO
2-rich systems.
Introduction
Nowadays Carbon capture and storage (CCS) is being studied as one of the most promising process to reduce the CO
2emissions from burning fossil fuels. Several research projects have been focused in improving the safety and reducing both the costs and the environmental risks associated with CCS operations (Li et al. 2011). From a thermodynamic point of view, an accurate description of thermophysical properties is crucial to the design of capture, transport and storage of CO
2. Therefore, the research and the assessment of new equation of state (EoS) models are important for the deployment of more accurate models.
In the oil and gas industry, with regard to the upstream, there is no interest in looking for more accurate EoSs, because in reservoir simulations the cubic EoSs provide enough accuracy due to uncertainties in the geology and the flow in porous media. Likewise, considering the computational time, thanks to their simplicity, cubic EoSs are scarcely replaced by more complex PVT models in reservoir simulations.
However in the downstream industry and CCS, more complex thermodynamic models can be applied in
order to increase the accuracy of PVT modeling: for example non-cubic EoSs can provide a better
density description at high pressures and high temperatures (HPHT) (Yan et al. 2015).
This paper focuses on one of the latest versions of the family of SAFT EoS, the SAFT-VR Mie. The adequacy of this EoS to describe transport properties like viscosity and interfacial tension (IFT), when coupled with appropriated models, has been investigated with results showing a very good agreement predictions and experimental data. The investigated systems include five binaries (CO
2/O
2, CO
2/N
2, CO
2/Ar, CO
2/n-C
4and CO
2/n-C
10) and two multicomponent mixtures (90%CO
2/ 5%O
2/ 2%Ar / 3%N
2and 90%CO
2/ 6%n-C
4/ 4%n-C
10) in a wide range of pressure and temperature conditions.
Model Description
Equation of state
The SAFT-VR Mie EoS proposed by Lafitte et al. 2013 is one of the latest versions of the SAFT equations of state. The SAFT-VR Mie is featured by the use of the Mie potential to describe the attraction-repulsion interactions between the segments which build the molecules. The Mie potential is defined as:
r a r ar
r r r
u
a r a r
r Mie
)
( (Eq. 1)
where r is radial distance, ζ the temperature-independent segment diameter, ε the potential depth; λ
rand λ
aare the repulsive and attractive ranges, respectively. The Mie parameters for the substances studied in this work are reported in the Table 1.
Table 1: Molecular parameters of the compounds used in this work (Lafitte et al. 2013)
Substance ms σ/Å (ε/k)/K λr λa
Carbon dioxide 1.5000 3.1916 231.88 27.557 5.1646
n-butane 1.8514 4.0887 273.64 13.650 6
n-decane 2.9976 4.5890 400.79 18.885 6
Nitrogen 1.4214 3.1760 72.438 9.8749 6
Oxygen 1.4283 2.9671 81.476 8.9218 6
Argon 1.0000 3.4038 117.84 12.085 6
The SAFT-VR Mie EoS can be expressed in terms of the reduced Helmholtz energy as the summation of several contributions:
ASSOC CHAIN
MONO IDEAL
a a
a NkT a
a A (Eq. 2)
where a
IDEALis the ideal gas contribution, a
MONOthe monomer contribution, a
CHAINthe contribution due to chain formation and a
ASSOCthe association contribution. The expressions of each contribution can be found in the studies of Lafitte et al. 2013.
Regarding to the phase equilibria, the thermodynamic criterion for phase equilibrium is the equality of
chemical potentials of each component in all co-existing phases. Using the SAFT EoS, the chemical
potential can be derived as µ=(∂A/∂N)
V,Tand then used to calculate the fugacity of each component
(Chapoy et al. 2013). In addition to the pure molecular parameters, temperature independent binary
interaction parameters k
ijhave been adjusted against solubility data (database Knapp et al. 1982) to
improve the description of vapor-liquid equilibria. The optimum k
ijparameters used in Eq. 3 are
reported in Table 2. The VLE predictions using SAFT-VR Mie with the adjusted k
ijare presented in the
Fig. 1. The percentage absolute average deviation (%AAD) between the predicted saturation pressure
and vapor composition and experimental data of Nagarajan are 2.8%Psat and 0.9%y.
i j ijj i ij
ij
k
33 3
1
(Eq. 3)
Table 2: Binary interaction parameters between O
2, Ar, N
2and CO
2, and between CO
2, n-butane and n- decane for the SAFT-VR Mie used in this work.
N2 O2 Ar CO2 n-C4 n-C10
N2 0 -0.009 -0.017 -0.127 - -
O2 -0.009 0 -0.023 -0.033 - -
Ar -0.017 -0.023 0 -0.014 - -
CO2 -0.127 -0.033 -0.014 0 0.053 0.057
n-C4 - - - 0.053 0 -0.004
n-C10 - - - 0.057 -0.004 0
0 2 4 6 8 10 12 14 16 18 20
0 0.2 0.4 0.6 0.8 1
P / M P a
x
CO2Fig. 1: Predicted VLE by SAFT-VR Mie for CO2-nC4 (grey) (Nagarajan 1985) and CO2-nC10 (black) (Nagarajan 1986).
Symbols: () T=344.3 and (▲) T=377.6K.
Viscosity model
TRAPP (TRAnsport Properties Prediction) is a predictive model based on the extended corresponding states theory (ECS) (K. C. M. and K. E. Gubbins 1976; Hanley and Cohen 1976), used to estimate the viscosity and thermal conductivity of pure fluids and their mixtures over the entire phase range. In the original TRAPP, propane was employed as reference fluid (Ely and Hanley 1983), although several reference fluids can be selected (Huber et. 1992). Accordingly, the residual viscosity of the mixture at a corresponding state point (T
0and ρ
0) is given by:
ENSKOG R
R m m
m
F
T
( , )
0[
0] (Eq. 4)
where η
m0is the viscosity of the mixture at low pressure evaluated by Herning and Zipperer (1936) approximation. The term η
R-η
R0is the residual viscosity of the reference fluid which is calculated as
(Younglove and Ely 1987):
1 3 ,
5 . 0 0 2 1 . 0 0 1
0
G exp[ G (
rR1 ) G ] G
R
R
(Eq. 5)
where η
Ris the real viscosity at T
0and ρ
0of the reference fluid, η
R0is the viscosity at low pressure and T
0, and ρ
r,Ris the reduced density calculated as ρ
r,R= ρ
0/ ρ
c,R. The corresponding state (T
0and ρ
0) is calculated as T
0=T/f
mand ρ
0=ρh
m. G
1, G
2and G
3are parameters that can be calculated as:
E E T
G
1 exp
1
2(Eq. 6)
5 . 1 4 3
2
E E T
G (Eq. 7)
2 7 6
5
3
E E T E T
G (Eq. 8)
where the E
iparameters have been correlated for propane as reference fluid and are shown in Table 3.
Table 3: E
iparameter TRAPP Model (Poling et al. 2001).
E1 -14.113294896
E2 968.22940153
E3 13.686545032
E4 -12511.628378
E5 0.01168910864
E6 43.527109444
E7 7659.4543472
In order to calculate the terms F
ηmand ∆η
ENSKOG, the following mixing rules were applied:
i j
ij j i
m
y y h
h (Eq. 9)
ij
i j
ij j i m
m
h y y h f
f Eq. 10)
8 ) ( )
(
i 1/3 j 1/3ij
h
h h
(Eq. 11)
2 /
)
1(
i jij
f f
f (Eq. 12)
where y
iis the mole fraction of component i; f
iand h
iare functions of the critical parameters and acentric factor defined as:
)]
ln 7498 . 0 05203 .
0 )(
( 1
[
R rR C
C
i
T
T
f T (Eq. 13)
)]
ln 2822 . 0 1436 . 0 )(
( 1
[
R rC R C C
R C
i
T
Z
h Z
(Eq. 14)
The viscosity dimensional scaling factor can be calculated as:
3 / 4 2 / 1 2
2 /
1
( ) ( ) ( )
)
(
ij iji j
ij j i m
R
m
M h y y f M h
F
(Eq. 15)
with
j i
j i
ij
M M
M M M
2
(Eq. 16) where M
idenotes the molecular weight.
The expression ∆η
ENSKOGis the residual viscosity which takes into account the differences in the molecules size based on the hard sphere assumption (Poling et al. 2001):
ENSKOG x ENSKOG m
ENSKOG
(Eq. 17)
with
i i
i hs
ij ij
i j
ij j i ENSKOG
m
y y g Y
2 6 0(Eq. 18)
where the α parameter is α=9.725x10
-7and ζ
iis the hard-sphere diameter calculated as ζ
i=4.771 h
i1/3. The radial distribution function of the hard-sphere fluid is given by the following equation (Gubbins and Gray 1972):
3 22
2
1
2 1
3 1
1
ij ij
hs
g
ij
(Eq. 19)
where
3 2
2
kk k
k k
k
ij j i
ij
y
y
(Eq. 20)
3
6
k ky
k
(Eq. 21)
To calculate the expression Σβ
iY
iit is necessary to solve the linear system of equations of the form:
i i
i
ij
Y
B (Eq. 22)
with
ij ijhsj i
j j
j i
i
g
M M y M y
Y
315
1 18
(Eq. 23)
jkk i ij k i
k i k
k ik
ik k i
ij
M
M M
M M
M M y g
y
B
3
2 3
1 5 2
2
0
(Eq. 24)
where δ
ijis the Kronecker function which is a delta function (i.e., δ
ij=1 if i=j and δ
ij=0 in other cases).
Interfacial tension model
The density gradient theory (DGT) has been used to compute the interfacial tension. This model is based
on the square gradient of van der Waals (Rowlinson 1979) and on the reformulation of Cahn and Hilliard 1958 to compute interfacial tension values from bulk phase properties such as density and composition. The DGT, when coupled with appropriate thermodynamic models, has been successfully applied in the prediction of interfacial properties of a wide class of systems and interfaces. The reader is referred to the study of Pereira et al. (2015) and references within for more details. In this work, the SAFT-VR Mie EoS was used to estimate the bulk equilibrium of the investigated systems and DGT used to predict interfacial properties. The main equations within the DGT framework are given below.
In summary, by applying the minimization criterion of the Helmholtz energy to planar interfaces, the interfacial tension values with respects to the density of a reference component is given by Miqueu et al. 2004 and 2005:
L ref
V
ref i j
ref ref
j ref
i
ij
d
d d d c d IFT
2 (Eq. 25)
where ρ
Lrefand ρ
Vrefare the bulk phase densities and the ref subscript denotes the reference component of the mixture. ∆Ω is the variation of the grand thermodynamic potential which is related to the Helmholtz free energy by the following equation:
p
f
i i i
0 (Eq. 26)
where f
0is the Helmholtz free energy density of the homogeneous fluid at local density, µ
iare the chemical potential of each component and p is the pressure at equilibrium. The methodology described by Miqueu et al. 2004 was here followed for determining the density distribution of each component across the interface.
In Eq. 25, c
ijis the cross influence parameter and the mixing rule used based on the geometric mean of the pure component influence parameters (Carey 1979) and it is given by
ij
i jij
c c
c 1 (Eq. 27)
where β
ijis the binary interaction parameter and c
iand c
jare the pure component influence parameters.
In this work, the binary parameter β
ijhas been fixed to 0, making the calculation of the interfacial tension of the systems investigated fully predictive. The influence parameters c
iand c
jcan be derived from theoretical expressions (Freeman and McDonald 1973). Nevertheless, the parameters are generally correlated using Surface Tension (ST) data from pure substances, by rewriting Eq. 25 for c
ias follows:
2
0
2 1
f p d c
LST
V
i
(Eq. 28)
There are different approaches for calculating the influence parameters of each pure component (Miqueu et al. 2004). In this work, c
ihas been taken as a constant value calculated from ST data far from the critical point (Lafitte et al. 2010). Hence, the influence parameters used were adjusted against ST data (NIST webbook 2013) of the pure components at a reduced temperature t
r=0.7. The parameters used are reported in Table 4.
Table 4: Density gradient theory influence parameters used in Eq. 26.
Substance ci / J·m5·mol-2
CO2 2.37 x 10-20 n-butane 1.68 x 10-19 n-decane 9.51 x 10-19
Results
Viscosity
The viscosity of CO
2systems was calculated using the TRAPP model and the densities computed from the SAFT-VR Mie EoS. A literature review of the available experimental data on the viscosity of some mixtures of relevance for CCS operations was carried and the results listed in Table 5. As can be seen in Table 5, experimental data on this property are available in a broad range of temperatures but, in most studies, pressure is limited to either atmospheric pressure or to a maximum pressure of 2.6MPa. Only in the study of Chapoy et al. 2013 this property has been investigated for temperatures in the range (273- 423) K and pressures up to 150MPa. The viscosity studies were conducted in the single phase region, i.e. Gas (G) phase at low pressures, Liquid (L) above saturation pressure or in the supercritical (SC) region, as described in Table 5.
Table 5: Literature experimental data for the viscosity of CO
2-mixtures with N
2, O
2, and Ar.
Source System Phase T/K P/MPa N Uncertainty
Kestin and Leidenfrost (1959) CO2/N2, CO2/N2 G 293 0.1-2.17 28 ±0.05%
Kestin et al. (1966) CO2/N2, CO2/Ar G 293-304 0.1-2.6 83 ±0.1%
Gururaja et al. (1967) CO2/N2, CO2/O2 G 298 0.1 17
Kestin and Ro (1974) CO2/N2, CO2/Ar, CO2/N2/Ar G 297-773 0.1 64 ±0.15%
Kestin et al. (1977) CO2/O2 G 298-674 0.1 10 ±0.3%
Hobley et al. (1989) CO2/Ar G 301-521 0.1 24 <0.7%
Chapoy et al. (2013) CO2/N2/O2/Ar G/L/ SC 273-423 1.5-150 38
Total 264
The modeling results for the viscosity of the CO
2-rich systems are shown in Table 6, along with the percentage absolute average deviation (%AAD) between predicted and experimental viscosity values for each studied system. A total of 264 experimental points were considered and an overall %AAD of 2.68
% was calculated for the viscosity values predicted with the TRAPP+SAFT-VR Mie model. This value represents half of the deviation reported by Huber (1996) and Poling et al. (2001) and obtained with the TRAPP model combined with a modified BWR-EoS. Nonetheless, the largest deviations with the TRAPP+SAFT-VR Mie model were obtained at high pressures and low temperatures, as depicted in Fig. 2, with a maximum deviation of 9.3% for the multicomponent system (90%CO
2, 5%O
2, 2%Ar and 3%N
2).
Table 6: Absolute average deviation of the viscosity calculations.
System xCO2 N T/K P/MPa %AAD
CO2/N2 0.19-0.90 95 293-773 0.1-2.6 3,12
CO2/Ar 0.22-0.92 96 293-773 0.1-2.6 1,28
CO2/O2 0.19-0.92 19 298-674 0.1 3,25
CO2/N2/Ar 0.25-0.53 16 297-673 0.1 1,39
CO2/N2/O2/Ar 0.9 38 273-423 1.5-150 5.15
Total 2.68%
0 50 100 150 200 250
0 50 100 150
Viscosity,η/ µPa·s
P/MPa
273K
298K
323K
373K
423K
Fig. 2: Experimental (Chapoy et al. 2013) and predicted viscosity of a CO2-rich system (90%CO2, 5%O2, 2%Ar and 3%N2).
Interfacial tension
The saturated densities and IFT of CO
2-rich systems were modeled by coupling the DGT with the
SAFT-VR Mie EoS. The results for the binaries CO
2-n-C
4and CO
2-n-C
10and the ternary CO
2-n-C
4-n-
C
10are plotted in Fig. 3 and Fig. 4, respectively. As can be seen in Fig. 3 and Fig. 4, the DGT+SAFT-
VR Mie predictions are in excellent agreement with the measured saturated density and IFT data of the
systems considered. Largest deviations were observed close to the critical point, where the SAFT-VR
Mie fails to capture the impact of pressure on the density of the bulk phases. Nonetheless, as can be seen
in Figure 5, the maximum absolute deviation between predicted and measured IFT data was of
0.26mNm
-1. In general, the IFT values for the CO
2-n-C
4system and the ternary system are slightly
overestimated, whereas an inverse behavior was observed for the CO
2-n-C
10system.
0 2 4 6 8 10 12 14 16
0 0.2 0.4 0.6 0.8 1
IFT / mN·m-1
xCO2 0
2 4 6 8 10 12 14 16 18
0 200 400 600 800
P / MPa
ρ / kg·m-3
(a) (b)
Fig. 3: Predicted density (a) and IFT (b) of CO2-nC4 (black) (Nagarajan 1985) [ref] and CO2-nC10 (grey) (Nagarajan 1986).
Symbols: () T=344.3 and (▲) T=377.6K.
0 2 4 6 8 10 12 14
0 200 400 600 800
P / MPa
ρ / kg·m-3
0 2 4 6 8 10 12 14 16
0.2 0.4 0.6 0.8 1
IFT / mN·m-1
xCO2 (a) (b)
Figure 4: Predicted density (a) and IFT (b) of CO2-nC4-nC10 mixture at 344.3K (Nagarajan 1990).
-0,3 -0,2 -0,1 0 0,1 0,2 0,3
0 5 10 15 20
IFT Error / mN·m-1
P / MPa
CO2-nC4 CO2-nC10 CO2-nC4-nC10
Fig. 5: Absolute deviation between prediction and experimental IFT data (IFTerror=IFTEXP–IFTCALC).
In addition, the density profiles through the interface were computed with the DGT approach for the ternary system (90%CO
2, 6%n-butane and 4%n-decane) and the results plotted in Fig. 5. As depicted in Fig. 5(a), the density profiles show a local enrichment of the interface with CO
2molecules as a peak was noticeable in the density profile of carbon dioxide. The location of peak closer to the vapor side of the interface suggests that the absorption of CO
2occurred at the surface of the liquid hydrocarbon phase.
Furthermore, as can be seen in Fig. 5(b), the pressure increase lead to an increase of the interface thickness in this mixture and to a reduction of the local accumulation on light components, such as CO
2. Comparable results were obtained for the binary mixtures and the results are in agreement with those obtained by Miqueu et al. 2005.
2 3 4 5 6 7 8 9 10 11
-4 -2 0 2 4
ρ / kmol·m-3
z / nm
8.3 MPa
4.6 MPa 6.3 MPa
0 1 2 3 4 5 6 7 8 9
0 2 4 6 8
ρ / k m o l· m
-3z / nm
(a) (b)
Fig. 5
:
(a) Density profiles calculated for each component in the CO2-nC4-nC10 mixture at 344.3K and 9.31MPa: CO2-(blue), n-butane (green) and n-decane (red). Total interface length of 7.36 nm. (b) Density profiles of carbon dioxide calculated for the CO2-nC10mixture at 344.3K at different pressures against the results of molecular simulation at same conditions (Müller and Mejía 2009).